Welcome! Here, we showcase the latest advancements and insights in data science and artificial intelligence. Dive into our curated collection of trends, topics, and discussions, and stay at the forefront of Data Science / AI developments. Join us in exploring the cutting-edge of technology and innovation.
Posted on: September 13, 2024
Almost every industry, company or organization today is going through some form of digital transformation that results in greater and greater quantities of both structured and unstructured data. The biggest challenge that companies have is turning their unusable, unstructured data... ...moreAlmost every industry, company or organization today is going through some form of digital transformation that results in greater and greater quantities of both structured and unstructured data. The biggest challenge that companies have is turning their unusable, unstructured data into useful insights that can help them make data driven decisions, create operational efficiencies, value improvement, and overall competitive advantage. Any industry can benefit from text analytics and sentiment analysis, as all industries collect data and require that it is transformed into actionable, tangible intelligence that can be applied to drive change. ...less
Posted on: January 31, 2024
Talk2Arxiv is an open-source RAG (Retrieval-Augmented Generation) system specially built for academic paper PDFs. Powered by talk2arxiv-server. Prepend any arxiv.org link with \'talk2\' to load the paper into a responsive RAG chat application (e.g. www.arxiv.org/pdf/1706.03762.pdf -> www.talk2arxiv.org/pdf/1706.03762.pdf) ...moreTalk2Arxiv is an open-source RAG (Retrieval-Augmented Generation) system specially built for academic paper PDFs. Powered by talk2arxiv-server.
Prepend any arxiv.org link with \'talk2\' to load the paper into a responsive RAG chat application (e.g. www.arxiv.org/pdf/1706.03762.pdf -> www.talk2arxiv.org/pdf/1706.03762.pdf) ...less
Posted on: January 19, 2024
Imagine a ChatGPT-like interface over all your structured (database) and unstructured data. Ideally, you want to ask a question to an AI bot, and it should be able to run multiple parallel queries on your database, look up relevant information... ...moreImagine a ChatGPT-like interface over all your structured (database) and unstructured data. Ideally, you want to ask a question to an AI bot, and it should be able to run multiple parallel queries on your database, look up relevant information from your docs, collate all the relevant information, and create a coherent response.
As an example - it should be able to respond intelligently to the request - Pull the top 10 customers by usage, closed by the sales rep John Doe, and draft a \"thank you\" email to each of them.
Your AI agent/ Custom LLM must do multiple things to satisfy this request.
- pull all the customers closed by sales rep John Doe from Salesforce
- run a query on Snowflake to get the usage data per customer
- find the thank you email template from your Google Docs folder
- draft emails to all of these customers
Systems that are capable of tasks that are moderately complex, like this one, really improve employee productivity significantly. Organizations that embrace AI in this way will easily outperform their competition.
So here is how you would build this
Data pipelines and connectors - You need connectors to systems like Salesforce, snowflake, and folders.
Chat query orchestrator - you need to be able to parse incoming chat queries, fire multiple calls to an LLM, generate SQL queries, or craft API requests to your vector databases (doc retrievers)
Doc retrievers/vector stores - You must embed and create a vector store for all your shared docs.
Final response creator - A final LLM that will generate the final response based on all the data. In the example above, it would generate the final emails from the email template and the data generated by the SQL queries.
Building these systems from scratch is doable, and there are several open-source libraries you can use to do so. However, the challenge will be iterating on the solution to make sure that the accuracy is high enough to be usable in production.
Several systems have to be tuned, and multiple iterations need to evaluate the system before putting this in production.
Our platform, Abacus AI, can help build, test, and launch a system like this in a matter of days. You can set up the various components quickly and then create multiple iterations and evaluate them across multiple LLMs and configurations.
Finally, you can set up all the pipelines and push the system to production in a matter of days. We offer several open-source fine tunes, which means you won\'t be paying for an expensive LLM if you don\'t have to! ...less
Posted on: January 6, 2024
Prompt engineering is a relatively new discipline for developing and optimizing prompts to efficiently use language models (LMs) for a wide variety of applications and research topics. Prompt engineering skills help to better understand the capabilities and limitations of large... ...morePrompt engineering is a relatively new discipline for developing and optimizing prompts to efficiently use language models (LMs) for a wide variety of applications and research topics. Prompt engineering skills help to better understand the capabilities and limitations of large language models (LLMs). Researchers use prompt engineering to improve the capacity of LLMs on a wide range of common and complex tasks such as question answering and arithmetic reasoning. Developers use prompt engineering to design robust and effective prompting techniques that interface with LLMs and other tools.
Motivated by the high interest in developing with LLMs, we have created this new prompt engineering guide that contains all the latest papers, learning guides, lectures, references, and tools related to prompt engineering for LLMs. ...less
Posted on: January 5, 2024
Recently-proposed large language models (LLMs) such as Google Gemini are structured and trained in a manner that maximizes efficiency and boosts training stability. But, what common tricks are used to achieve these efficiency/stability benefits? TL;DR: Making LLMs more efficient—both during... ...moreRecently-proposed large language models (LLMs) such as Google Gemini are structured and trained in a manner that maximizes efficiency and boosts training stability. But, what common tricks are used to achieve these efficiency/stability benefits?
TL;DR: Making LLMs more efficient—both during training and inference—boils down to three basic categories of techniques:
1. Modifying the model’s architecture.
2. Modifying the model’s weights.
3. Modifying the training/optimization process.
This post covers several techniques that fall under either of these categories and are commonly used by a variety of recent, state-of-the-art LLMs.
(1) Modifying the architecture. To make an LLM more efficient/stable, we can first try to modify the model’s architecture. Notable examples of this include:
- Flash Attention: a hardware-aware efficient attention implementation that can drastically speed up training (used by numerous LLMs like MPT-7/30B and Falcon).
- Flash Decoding: an extension of flash attention that improves attention efficiency in the inference/decoding stage (in addition to training).
- Multi-Query Attention: an efficient attention implementation that shares key and value vectors between attention heads (used by PaLM and LLaMA-2).
(2) Tweaking the weights. Beyond changing the architecture itself, we can also modify the model’s weights without changing the underlying architecture. More specifically, we can quantize the model’s weights into a lower precision representation, either during training or inference (or both). Example of this include:
- Post-training Quantization: we quantize the LLM’s weights to a lower precision representation prior to inference and perform inference in low precision (commonly used to save costs when hosting/deploying LLMs).
- Low Precision LayerNorm: performs layer normalization in low precision to improve training efficiency without impacting model quality (used by MPT-7/30B).
- Quantized training: we quantize the LLM’s weights throughout the entire training process to yield a speedup. The most common example of this is Automatic Mixed Precision (AMP) training, which is widely used in deep learning.
(3) Changing the optimization process. Finally, we can also explore changes to the model’s underlying optimization process. Typically, these changes are a bit difficult to explore—training LLMs is incredibly expensive and we don’t want to pay for training runs that yield poor results. Nonetheless, several interesting techniques have been explored:
- Lion Optimizer: a new adaptive optimizer (i.e., this would replace another optimizer like AdamW) that has been shown (by LLMs like MPT-7B) to improve the stability of LLM training.
- Adaptive Sequence Length: models like MPT-7/30B and Orca have shown us that we can speed up the LLM training process by first training on shorter / less complex examples, then training over longer / more complex examples.
- LoRA: parameter efficient finetuning techniques like LoRA, LLaMA Adapter, and many more achieve impressive results by training efficient, low-rank parameterizations of LLMs. Arguably, these approaches modify more than just the optimization process (i.e., the model is changed as well), but they are nonetheless incredibly effective. ...less
Posted on: January 5, 2024
This survey is designed for both financial economists interested in grasping machine learning tools, as well as for statisticians and machine learners seeking interesting financial contexts where advanced methods may be deployed. ...moreThis survey is designed for both financial economists interested in grasping machine learning tools, as well as for statisticians and machine learners seeking interesting financial contexts where advanced methods may be deployed. ...less
Posted on: January 4, 2024
You’ll learn the following: (i) Query expansion using an LLM to rewrite and improve a query, by either generating either additional relevant queries or a hypothetical answer to the query. (ii) Reranking using a cross-encoder - a model trained to... ...moreYou’ll learn the following:
(i) Query expansion using an LLM to rewrite and improve a query, by either generating either additional relevant queries or a hypothetical answer to the query.
(ii) Reranking using a cross-encoder - a model trained to measure similarity between two inputs presented simultaneously. Reranking reorders retrieved documents based on the cross-encoder similarity measure.
(iii) Constructing and training an Embedding Adaptor, which is a model that adapts the embedding values to be more relevant to your use case.
Each of these techniques can help you build much better RAG systems. ...less
Posted on: January 3, 2024
1. Dario Amodei, Anthropic 2. Greg Brockman, OpenAI 3. Clem Delangue, Hugging Face 4. Aidan Gomez, Cohere 5. Robin Li, Baidu 6. Mira Murati, OpenAI 7. Sarah Nagy, Seek AI 8. Naveen Rao, Databricks 9. Rahul Roy-Chowdhury, Grammarly 10. Ovetta... ...more1. Dario Amodei, Anthropic
2. Greg Brockman, OpenAI
3. Clem Delangue, Hugging Face
4. Aidan Gomez, Cohere
5. Robin Li, Baidu
6. Mira Murati, OpenAI
7. Sarah Nagy, Seek AI
8. Naveen Rao, Databricks
9. Rahul Roy-Chowdhury, Grammarly
10. Ovetta Sampson, Google
11. Silvio Savarese, Salesforce
12. Kevin Scott, Microsoft
13. Karén Simonyan, Inflection AI
14. Mai-Lan Tomsen Bukovec, Amazon Web Services
15. Diya Wynn, Amazon Web Services
You can also find this list: The AI 100 2023: The top people in artificial intelligence (https://www.businessinsider.in/tech/news/the-ai-100-2023-the-top-people-in-artificial-intelligence/articleshow/104312755.cms) ...less
Posted on: January 3, 2024
Duet AI for Developers completes your code as you write, and generates whole code blocks or functions on demand. Code assistance is available in multiple IDEs such as Visual Studio Code, JetBrains IDEs (IntelliJ, PyCharm, GoLand, WebStorm, and more), Cloud... ...moreDuet AI for Developers completes your code as you write, and generates whole code blocks or functions on demand. Code assistance is available in multiple IDEs such as Visual Studio Code, JetBrains IDEs (IntelliJ, PyCharm, GoLand, WebStorm, and more), Cloud Workstations, Cloud Shell Editor, and supports 20+ programming languages, including Go, Java, Javascript, Python, and SQL. ...less
Posted on: January 2, 2024
This guide shares strategies and tactics for getting better results from large language models (sometimes referred to as GPT models) like GPT-4. The methods described here can sometimes be deployed in combination for greater effect. We encourage experimentation to find... ...moreThis guide shares strategies and tactics for getting better results from large language models (sometimes referred to as GPT models) like GPT-4. The methods described here can sometimes be deployed in combination for greater effect. We encourage experimentation to find the methods that work best for you.
Some of the examples demonstrated here currently work only with our most capable model, gpt-4. In general, if you find that a model fails at a task and a more capable model is available, it\'s often worth trying again with the more capable model. ...less
Posted on: January 2, 2024
The LLM course is divided into three parts: 1. LLM Fundamentals covers essential knowledge about mathematics, Python, and neural networks. 2.The LLM Scientist focuses on learning how to build the best possible LLMs using the latest techniques 3. The LLM... ...moreThe LLM course is divided into three parts:
1. LLM Fundamentals covers essential knowledge about mathematics, Python, and neural networks.
2.The LLM Scientist focuses on learning how to build the best possible LLMs using the latest techniques
3. The LLM Engineer focuses on how to create LLM-based solutions and deploy them. ...less
Posted on: December 30, 2023
1. Pythia: A Suite for Analyzing Large Language Models Across Training and Scaling 2. Llama 2: Open Foundation and Fine-Tuned Chat Models 3. QLoRA: Efficient Finetuning of Quantized LLMs 4. BloombergGPT: A Large Language Model for Finance 5. Direct Preference... ...more1. Pythia: A Suite for Analyzing Large Language Models Across Training and Scaling
2. Llama 2: Open Foundation and Fine-Tuned Chat Models
3. QLoRA: Efficient Finetuning of Quantized LLMs
4. BloombergGPT: A Large Language Model for Finance
5. Direct Preference Optimization: Your Language Model is Secretly a Reward Model
6. Mistral 7B
7. Orca 2: Teaching Small Language Models How to Reason
8. ConvNets Match Vision Transformers at Scale
9. Segment Anything
10. Align your Latents: High-Resolution Video Synthesis with Latent Diffusion Models ...less
Posted on: December 29, 2023
1. AI smart glasses become a thing 2. ChatGPT won\'t be to AI assistant what Google is to search 3. So long LLMs, hello LMMs (Large Multimodals Models) 4. Small is Beautiful - Small Language Models (SLMs) 5. An open... ...more1. AI smart glasses become a thing
2. ChatGPT won\'t be to AI assistant what Google is to search
3. So long LLMs, hello LMMs (Large Multimodals Models)
4. Small is Beautiful - Small Language Models (SLMs)
5. An open model beats GPT-4, yet the open vs closed debate progressively fades ...less
Posted on: December 29, 2023
Reasoning is going to be a super active area of focus in AI research in the coming months. If you want to catch up, this survey paper provides a comprehensive survey of seminal foundational models for reasoning. It highlights the... ...moreReasoning is going to be a super active area of focus in AI research in the coming months.
If you want to catch up, this survey paper provides a comprehensive survey of seminal foundational models for reasoning.
It highlights the latest advancements in various reasoning tasks, methods, benchmarks, and potential future directions.
It also discusses how other developments like multimodal learning, autonomous agents, and super alignment accelerate and extend reasoning research.
https://arxiv.org/abs/2312.11562v4 ...less
Posted on: December 28, 2023
Here are the insights from: - Anastasis Germanidis: New Tools to Tell New Stories - Sara Hooker: Prioritize Inclusion - Percy Liang: Transparency for Foundation Models - Sasha Luccioni: Respect for Human Creativity and Agency - Pelonomi Moiloa: Smaller Models... ...moreHere are the insights from:
- Anastasis Germanidis: New Tools to Tell New Stories
- Sara Hooker: Prioritize Inclusion
- Percy Liang: Transparency for Foundation Models
- Sasha Luccioni: Respect for Human Creativity and Agency
- Pelonomi Moiloa: Smaller Models That Learn More From Less Data
- Kevin Scott: Be Prepared for Another Year of Exponential Growth
What\'s not going to change:
- We need community
- People who know how to use AI tools are more productive
- AI needs good data to function well ...less